Solution synthesis of a new thermoelectric Zn1+ xSb nanophase and its structure determination using automated electron diffraction tomography

Article abstract of DOI:10.1021/ja1035122

Engineering materials with specific physical properties have recently focused on the effect of nanoscopic inhomogeneities at the 10 nm scale. Such features are expected to scatter medium- and long-wavelength phonons thereby lowering the thermal conductivity of the system. Low thermal conductivity is a prerequisite for effective thermoelectric materials, and the challenge is to limit the transport of heat by phonons, without simultaneously decreasing charge transport. A solution-phase technique was devised for synthesis of thermoelectric Zn₄Sb₃ nanocrystals as a precursor for phase segregation into ZnSb and a new Zn-Sb intermetallic phase, Zn _1+δSb, in a peritectoid reaction. Our approach uses activated metal nanoparticles as precursors for the synthesis of this intermetallic compound. The small particle size of the reactants ensures minimum diffusion paths, low activation barriers, and low reaction temperatures, thereby eliminating solid-solid diffusion as the rate-limiting step in conventional bulk-scale solid-state synthesis. Both phases were identified and structurally characterized by automated electron diffraction tomography combined with precession electron diffraction. An ab initio structure solution based on electron diffraction data revealed two different phases. The new pseudo-hexagonal phase, Zn_1+δSb, was identified and classified within the structural diversity of the Zn-Sb phase diagram.

Full text of DOI:10.1021/ja1035122

Published on Web 06/29/2010
Solution Synthesis of a New Thermoelectric Zn1+xSb
Nanophase and Its Structure Determination Using Automated
Electron Diffraction Tomography
Christina S. Birkel,† Enrico Mugnaioli,‡ Tatiana Gorelik,‡ Ute Kolb,*,‡
Martin Pantho¨fer,† and Wolfgang Tremel*,†
Institut fu¨r Anorganische Chemie und Analytische Chemie der Johannes Gutenberg-UniVersita¨t,
Duesbergweg 10-14, D-55099 Mainz, Germany, and Institut fu¨r Physikalische Chemie der
Johannes Gutenberg-UniVersita¨t, Welderweg 11, D-55099 Mainz, Germany
Received April 29, 2010; E-mail: kolb@uni-mainz.de; tremel@uni-mainz.de
Abstract: Engineering materials with specific physical properties have recently focused on the effect of
nanoscopic inhomogeneities at the 10 nm scale. Such features are expected to scatter medium- and long-
wavelength phonons thereby lowering the thermal conductivity of the system. Low thermal conductivity is
a prerequisite for effective thermoelectric materials, and the challenge is to limit the transport of heat by
phonons, without simultaneously decreasing charge transport. A solution-phase technique was devised
for synthesis of thermoelectric “Zn4Sb3” nanocrystals as a precursor for phase segregation into ZnSb and
a new Zn-Sb intermetallic phase, Zn1+δSb, in a peritectoid reaction. Our approach uses activated metal
nanoparticles as precursors for the synthesis of this intermetallic compound. The small particle size of the
reactants ensures minimum diffusion paths, low activation barriers, and low reaction temperatures, thereby
eliminating solid-solid diffusion as the rate-limiting step in conventional bulk-scale solid-state synthesis.
Both phases were identified and structurally characterized by automated electron diffraction tomography
combined with precession electron diffraction. An ab initio structure solution based on electron diffraction
data revealed two different phases. The new pseudo-hexagonal phase, Zn1+δSb, was identified and classified
within the structural diversity of the Zn-Sb phase diagram.
of “Zn4Sb3” have been studied for decades.5,6,7,8-16 Common
to all four known phases in the phase diagram is a framework
Introduction
Thermoelectric materials allow the conversion of a thermal
gradient into a gradient of the electrical potential and may play
a vital role in the search for alternative energy technologies.1,2
The thermoelectric performance of a material is expressed by a
dimensionless figure of merit, ZT ) S2T/Fκ, where S is the
thermopower; F is the electrical resistivity; κ is the thermal
conductivity; and T is the temperature. Heat is carried by both
electrons (κe) and phonons (κph),3 thus the total thermal
conductivity is the sum of the lattice thermal conductivity and
the electronic thermal conductivity. While the latter can be
estimatedfromtheelectricalconductivityusingtheWiedemann-Franz
law,6 the lattice thermal conductivity is closely related to the
crystal structure.
4-
of isolated Sb3- anions and Sb2 dumbbells, the latter being
centered in a metaprism formed by six Zn cations. Apart from
the high-temperature γ-phase, all low-temperature phases exhibit
occupational disorder in the Zn partial structure and occupation
of interstitials with decreasing level of order upon decreasing
(5) Caillat, T.; Fleurial, J. P.; Borshchevsky, A. J. Phys. Chem. Solids
1997, 58, 1119–1125.
(6) Elliott, S. The Physics and Chemistry of Solids; Wiley and Sons:
Chichester, 1998.
(7) Kim, S. G.; Mazin, I. I.; Singh, D. J. Phys. ReV. B 1998, 57, 6199–
6203.
(8) Mayer, H. W.; Mikhail, I.; Schubert, K. J. Less-Common Met. 1978,
59, 43–52.
(9) Snyder, G. J.; Christensen, M.; Nishibori, E.; Caillat, T.; Iversen, B. B.
Nat. Mater. 2004, 3, 458–463.
Among thermoelectric materials, “Zn4Sb3” (more precisely
Zn3.9-xSb3) exhibits an outstanding figure of merit between 450
and 670 K with ZT ≈ 1.3 at 670 K, owing to its exceptionally
low thermal conductivity.4,5 Structure and precise stoichiometry
(10) Cargnoni, F.; Nishibori, E.; Rabiller, P.; Bertini, L.; Snyder, G. J.;
Christensen, M.; Gatti, C.; Iversen, B. B. Chem.sEur. J. 2004, 10,
3862–3870.
(11) Mozharivskyj, Y.; Pecharsky, A. O.; Bud‘ko, S.; Miller, G. J. Chem.
Mater. 2004, 16, 1580–1589.
† Institut fu¨r Anorganische Chemie und Analytische Chemie der Johannes
Gutenberg-Universita¨t.
(12) Nylen, J.; Andersson, M.; Lidin, S.; Ha¨ussermann, U. J. Am. Chem.
Soc. 2004, 126, 16306–16307.
‡ Institut fu¨r Physikalische Chemie der Johannes Gutenberg-Universita¨t.
(1) Goldsmid, H. J. Thermoelectric Refrigeration; Plenum Press: New
York, 1988.
(13) Mozharivskyj, Y.; Janssen, Y.; Harringa, J. L.; Kracher, A.; Tsokol,
A. O.; Miller, G. J. Chem. Mater. 2006, 18, 822–831.
(14) Nylen, J.; Lidin, S.; Andersson, M.; Iversen, B. B.; Liu, H.; Newman,
N.; Ha¨ussermann, U. Chem. Mater. 2007, 19, 834–838.
(15) Nylen, J.; Lidin, S.; Andersson, M.; Liu, H.; Newman, N.; Ha¨usser-
mann, U. J. Solid State Chem. 2007, 180, 2603–2615.
(16) Schweika, W.; Hermann, R. P.; Prager, M.; Persson, J.; Keppens, V.
Phys. ReV. Lett. 2007, 99, 125501.
(2) Wood, C. Rep. Prog. Phys. 1988, 51, 459–539.
(3) Ziman, J. M. Prinzipien der Festko¨rpertheorie; Verlag Harry Deutsch:
Zu¨rich und Frankfurt am Main, 1972.
(4) Bhattacharya, S.; Hermann, R. P.; Keppens, V.; Tritt, T. M.; Snyder,
G. J. Phys. ReV. B 2006, 74, 134108.
9
10.1021/ja1035122  2010 American Chemical Society
J. AM. CHEM. SOC. 2010, 132, 9881–9889 9881

A R T I C L E S
Birkel et al.
temperature. The amount of occupation of interstitials seems
to be sample dependent, thus “Zn4Sb3” has become a generic
term for the group of structurally comparable compounds
Zn6-δSb5 (δ ) 0.62, 0.76),13 Zn3.83Sb3,14 and Zn84Sb65.15
The proposed structures are variants of a Zn6Sb5 basis
structure with structural disorder due to additional Zn interstitials
and Zn vacancies9–12,14 and have been derived from compre-
hensive X-ray diffraction studies and analyses using maximum
entropy methods. Recent inelastic neutron scattering investiga-
tions relate low thermal conductivity not to a phonon glass but
to a soft localized vibration of dumbbell Sb2 units in the Sb
sublattice.4,16
The interdependency of transport properties in ZT makes the
engineering of thermoelectric materials a difficult task. However,
the physics of nanostructures allows varying S, F, and κ
independently, thus enhancing ZT under appropriately selected
conditions. The greatest gains in ZT have been achieved by
introducing interfaces such that phonons are preferentially
scattered relative to electrons. Different approaches to decrease
κph by reducing the mean free path of phonons while maintaining
the electronic properties of doped semiconductors have been
attempted: (i) alloying,17 (ii) nanostructuring,18,19 and (iii) guest
atom inclusion.20,21 Prominent examples of thermoelectric
materials based on the “phonon glass-electron crystal” concept22,23
are the skutterudites22 or “Zn4Sb3” studied here.
One strategy for creating the required scattering interfaces relies
on the synthesis of nanoparticles and multiphase composites mixed
on the nanometer scale. These nanostructured materials can be
formed as thin-film superlattices or as intimately mixed composite
structures. The first demonstration that a low-dimensional material
system could enhance thermoelectric performance was for a 2D
superlattice consisting of PbTe quantum wells, Pb1-xEuxTe barri-
ers,24 and (Bi,Sb)2Te3 superlattices.25,26 Epitaxial-type superlattice
structures are formed, and their sizes can be controlled by phase
separation of metastable ternary compounds into their correspond-
ing binaries.19,20,27,28
The following question thus arises: can we conceive other
materials, based on solidification and decomposition, which may
lead to enhanced ZT-values? Here, we present an extension of
the “solidification and decomposition” approach to thermoelec-
tric materials at the nanoscale using thermolabile “Zn4Sb3” as
the starting compound.
From a synthetic point of view, it is crucial to control the
nanostructure and morphology of these complex binary materials
and their performance in potential thermoelectric applications.
Controlling nanostructure and morphology is important, as they
are critical for the intended applications. Intermetallics are
typically made at high temperatures (usually >1000 °C) with
long annealing times (often days or weeks). Due to the
intrinsically metastable nature of nanoscale matter, we have
devised a low-temperature synthetic route for “Zn4Sb3” nano-
particles, where weakly stabilized Zn and Sb nanoparticles
aggregate to form binary nanocomposites29 reacting subse-
quently to form intermetallic nanocrystals. The synthesis can
be scaled up to yield bulk amounts of material. The novel
solution-based approach uses activated metal nanoparticles as
precursors for the synthesis of a Zn1+xSb nanophase, containing
ZnSb and a new binary phase Zn1+δSb, segregated from a nano-
Zn4Sb3 precursor in a peritectoid reaction.
Traditional single-crystal X-ray diffraction techniques cannot
be applied to nanostructures. Powder X-ray analysis meets
serious problems when the size of the crystals is below 50 nm
and different phases are present in the sample, all attributes of
the Zn-Sb system. Therefore, the crystallographic investigation
was performed using electron diffraction. In contrast to X-rays,
an electron beam can be focused on a very small area allowing
the selection of a single nanocrystal of only a few nanometers.
As scattering factors for electrons are significantly higher in
comparison to X-rays, diffraction information with a reasonable
signal-to-noise ratio can be collected even from such small
volumes. Two types of problems are usually attributed to
electron diffraction data, which diminishes its usability for
structure solution: electron diffraction data often miss a sufficient
number of reflections, and the intensities of the reflections are
strongly modified due to dynamical effects. To overcome these
problems and to develop a well-established routine for electron
diffraction data collection, techniques like automated diffraction
tomography (ADT)30,31 and precession electron diffraction
(PED)32-34 have been developed in the last years. By combining
these two techniques, a rich sampling of the reciprocal space
can be achieved, and thus almost complete quasi-kinematical
electron diffraction data sets can be collected. Mugnaioli et al.35
demonstrated that this method, comprising basically the 3D
reconstruction of the reciprocal space with quasi-kinematical
data, allows solving inorganic structures ab initio in one step.
This data collection/structure solution strategy was applied to
the Zn1+xSb nanophase containing a known ZnSb phase and a
hitherto unknown Zn1+δSb. Both phases were identified by
electron diffraction and high-resolution TEM imaging, and ab
initio structure determinations were performed using combined
ADT/PED.
(17) Kimura, Y.; Zama, A. Appl. Phys. Lett. 2006, 89, 103705.
(18) Hsu, K. F.; Loo, S.; Guo, F.; Chen, W.; Dyck, J. S.; Uher, C.; Hogan,
T. P.; Polychroniadis, E. K.; Kanatzidis, M. G. Science 2004, 303,
818–821.
(19) Poudeu, P. F. R.; D’Angelo, J.; Downey, A. D.; Short, J. L.; Hogan,
T. P.; Kanatzidis, M. G. Angew. Chem. 2006, 118, 3919-3923; Angew.
Chem., Int. Ed. 2006, 45, 3835-3839.
(20) Nolas, G. S.; Cohn, J. L.; Slack, G. A.; Schujman, S. B. Appl. Phys.
Lett. 1998, 73, 178–180.
Results and Discussion
(21) Keppens, V.; Mandrus, D.; Sales, B. C.; Chakoumakos, B. C.; Dai,
P.; Coldea, R.; Maple, M. B.; Gajewski, D. A.; Freeman, E. J.;
Bennington, S. Nature 1998, 395, 876–878.
Synthesis. Synthesis of the Zn1+xSb nanophase was achieved
through the reaction of “activated” zinc with “activated”
(22) Sales, B. C.; Mandrus, D.; Williams, R. K. Science 1996, 272, 1325–
1328.
(23) Nolas, G. S.; Morelli, D. T.; Tritt, T. M. Annu. ReV. Mater. Sci. 1999,
29, 89–116.
(29) Schlecht, S.; Erk, C.; Yosef, M. Inorg. Chem. 2006, 45, 1693–1697.
(30) Kolb, U.; Gorelik, T.; Kubel, C.; Otten, M. T.; Hubert, D. Ultrami-
croscopy 2007, 107, 507–513.
(24) Hicks, L. D.; Harman, T. C.; Sun, X.; Dresselhaus, M. S. Phys. ReV.
B 1996, 53, R10493.
(31) Kolb, U.; Gorelik, T.; Otten, M. T. Ultramicroscopy 2008, 108, 763–
772.
(25) Martin-Gonzalez, M.; Snyder, G. J.; Prieto, A. L.; Gronsky, R.; Sands,
T.; Stacy, A. M. Nano Lett. 2003, 3, 973–977.
(32) Vincent, R.; Midgley, P. A. Ultramicroscopy 1994, 53, 271–282.
(33) Own, C. S. System Design and Verification of the Precession Electron
Diffraction Technique, Ph.D. dissertation, Northwestern University,
2005.
(26) Caylor, J. C.; Coonley, K.; Stuart, J.; Colpitts, T.; Venkatasubramanian,
R. Appl. Phys. Lett. 2005, 87, 023105.
(27) Ikeda, T.; Ravi, V. A.; Snyder, G. J. J. Mater. Res. 2008, 9, 2538–
2546.
(34) Avilov, A.; Kuligin, K.; Nicolopoulos, S.; Nickolskiy, M.; Boulahya,
K.; Portillo, J.; Lepeshov, G.; Sobolev, B.; Collette, J. P.; Martin, N.;
Robins, A. C.; Fischione, P. Ultramicroscopy 2007, 107, 431–444.
(28) Ikeda, T.; Collins, L. A.; Ravi, V. A.; Gascoin, F. S.; Haile, S. M.;
Snyder, G. J. Chem. Mater. 2007, 19, 763–767.
9
9882 J. AM. CHEM. SOC. VOL. 132, NO. 28, 2010

Synthesis of a Thermoelectric Zn1+xSb Nanophase
A R T I C L E S
Figure 1. Time-dependent (ex situ) X-ray diffraction (experimental data: black circles), le Bail fits (red line), and corresponding difference plots (black line)
after reaction times of 60, 120, 180 (as prepared), and 225 min (after two washing cycles with dilute acetic acid).
antimony in the temperature range of 275-300 °C. “Activated”
reactants were produced by reduction of ZnCl2 and SbCl3 with
lithium triethylboronhydride Li[Et3BH] at 65 °C in tetrahydro-
furan. The black metal powders obtained were redispersed
repeatedly in dry tetrahydrofuran to remove residual lithium salts
and were subsequently dried in vacuo. Both Zn and Sb
nanoparticles are highly reactive in ambient air leading primarily
to the oxides. 1H NMR data of Sb nanoparticles kept in an inert
gas atmosphere show an intense signal according to surface bond
stibane (δ ≈ 1.05 ppm, c.f. Figure SI-1, Supporting Information)
which broadens when the particles are in contact with air.
The activated elements were heated to 300 °C at a rate of ca.
20 K/min in a polar strongly coordinating solvent (trioctylamine)
to prevent nanoparticle aggregation. Products were collected by
centrifugation, washed with ethanol, and subsequently dried in
vacuo. The syntheses had to be carried out using an excess of
zinc metal which is removed after reaction by repeated treatment
of the solid residue with dilute acetic acid. Prior to heating,
both Zn and Sb were crystalline, and the XRD pattern matched
that expected for a superposition of both elements. After several
minutes of heating at 300 °C, the intermetallic compound ZnSb
was formed; further heating increased its crystallinity. The
product is stable in air at least over the period of several months.
Combined TG/DTA of the product in the range from ambient
temperature to 450 °C showed an exothermic effect at 200 °C
corresponding to the peritectoid decomposition of Zn1+xSb,
forming ZnSb and Zn. Upon further heating, the melting of Zn
(ca. 410 °C) was observed, while only a minor weight loss
occurred.
temperature reactions, the energy required for the solid state
diffusion of the reactants would exceed the nucleation energy
of metastable phases. In contrast, reactions, in which solid state
diffusion plays only a minor role, are kinetically controlled, and
a phase may nucleate and grow until its growth exhausts the
supply of the reactants. The sequence of phases formed depends
upon the relative activation energies for nucleation of the various
compounds, and compounds in the equilibrium phase diagram
may be temporarily skipped if they have large activation energy
for nucleation. The size of the “activated” metal particles allows
for a reaction with minimum solid state diffusion, low activation
barriers, and low reaction temperatures, thereby eliminating
solid-solid diffusion as the rate-limiting step in conventional
bulk-scale solid-state synthesis. It follows the strategy proposed
by Johnson et al.36,37 on elementally modulated thin films that
was utilized later for the synthesis of fullerene-type chalcogenide
nanoparticles38 or in the so-called polyol synthesis for nano-
particles of noble metals39 and intermetallics.40 The nanosize
dimensions support low-temperature diffusion without the high-
temperature melting step that is necessary in traditional metal-
lurgy. Clearly, nanoscale intermixing of prereduced metal
particles significantly increases the reaction kinetics relative to
conventional high-temperature syntheses involving micrometer-
sized powders. Likewise, the solvent confines the crystallite size
to nanoscale dimensions, allowing atomic ordering to occur
rapidly without long-term annealing.
X-ray Diffraction. X-ray powder diffraction profiles recorded
from reaction intermediates after specified time intervals (Figure
(36) Novet, T.; Johnson, D. C. J. Am. Chem. Soc. 1991, 113, 3398–3403.
(37) Noh, M.; Thiel, J.; Johnson, D. C. Science 1995, 270, 1181–1184.
(38) Etzkorn, J.; Therese, H. A.; Rocker, F.; Zink, N.; Kolb, U.; Tremel,
W. AdV. Mater. 2005, 17, 2372–2375.
It was observed that the size of an activated metal particle is
crucial for the outcome of the reaction. In conventional high-
(39) Sun, Y. G.; Xia, Y. N. Science 2002, 298, 2176–2179.
(40) Leonard, B. M.; Bhuvanesh, N. S. P.; Schaak, R. E. J. Am. Chem.
Soc. 2005, 127, 7326–7327.
(35) Mugnaioli, E.; Gorelik, T.; Kolb, U. Ultramicroscopy 2009, 109, 758–
765.
9
J. AM. CHEM. SOC. VOL. 132, NO. 28, 2010 9883

A R T I C L E S
Birkel et al.
Figure 3. Small aggregates used for ADT data collections. (a) Data for
structural analysis of phase I were collected from the middle part of the
agglomerate. (b) Two data sets for structural analysis of phase II were
collected from crystals indicated by arrows. In the middle part of this
aggregate, diffraction patterns with interplanar distances of 11.2 Å were
observed, likely due to the presence of metallic Sb.
Figure 2. Overview TEM images of typical aggregates obtained after (a)
a reaction time of 1 h and (b) the final product after 3 h 45 min.
1) revealed that ZnSb was formed within 1 h. From the X-ray
powder data, it was evident that the primary product transformed
into a new phase with cell parameters geometrically related to
those of “Zn4Sb3” in the subsequent reaction steps. The data
collected after reaction intervals of 2 and 3 h indicated the
formation of a mixture of ZnSb and nominal Zn4Sb3, which
was the major product when the reaction was complete after
3 h 45 min. Upon further heating, a reformation of ZnSb due
to decomposition of Zn4Sb3 (mostly into Zn) was observed
(Figure SI-3, Supporting Information). If the amount of Zn was
less than a 9-fold excess, the major product of the reaction was
ZnSb instead of Zn4Sb3.
Figure 4. HRTEM images of Zn1+δSb nanoparticles viewed along (a) the
[001] and (b) the [011] directions.
All diffraction profiles included reflections of metallic Sb and
Zn (marked in Figure 1 by blue stars and magenta crosses,
respectively). The final product obtained after 225 min contains
only minor amounts of these impurities. The powder patterns
of all crystalline phases were refined using le Bail fits with lattice
cell parameters of ZnSb and Zn4Sb3.11 The fitting of the final
product was carried out using the crystallographic data of Zn4Sb3
and gave an almost straight difference curve indicating that there
are no further crystalline impurities. The crystallite size of the
Zn4Sb3 nanoparticles as determined from the refinement was
approximately 30 nm.
TEM Overview. In all TEM investigations, the samples
proved to be extremely stable; i.e., no alteration was observed
either in short-term or after hours of data acquisition. TEM
images as well as SEM images show that both the intermediate
and the final products consist of porous agglomerates built of
tens to hundreds of round particles (Figure 2, Figure SI-4,
Supporting Information) with diameters between 5 and 70 nm.
The mean particle size determined from TEM is therefore in
harmony with the crystallite size obtained from the powder
X-ray studies. Despite intensive sonication during TEM sample
preparation, individual crystals were never observed, and even
the smallest aggregates consisted of several crystals (Figure 3).
HRTEM images of the final Zn1+xSb nanophase demonstrated
that all particles but the smallest contain a crystalline core
surrounded by an amorphous shell of 3-10 nm (Figure 4, Figure
SI-5, Supporting Information). Although the sample crystallinity
judged from the powder X-ray profiles did not differ much for
the products after different reaction times, HRTEM imaging
showed that the final product (obtained after 3 h 45 min) has
significantly higher crystallinity.
described by any known phase. Obviously, the lattice parameters
of the new phase should be strictly geometrically related to
Zn6Sb5. The cell volume of the new phase, calculated by high-
order Laue zones (Figure SI-5, Supporting Information), is
approximately 1630 Å3. EDX analyses of multiple regions of
the samples confirm that Zn and Sb are present in an ap-
proximate 1:1 atomic ratio with a slight excess of Zn (53% Zn,
47% Sb; Figure SI-6, Supporting Information).
Electron Crystallography by ADT. For structural investiga-
tions, ADT data were collected from prominent crystals in
relatively small aggregates (Figure 3) using an electron beam
with 50 nm diameter (for details, see refs 30 and 31). 3D
reconstruction of the reciprocal space for different crystals
showed unambiguously that the sample consisted at least of two
different phases. While the first phase had an orthorhombic cell
(phase I), the second one showed a lattice with pseudo-
hexagonal metric (phase II). Approximate average compositions
determined from EDX spectra did not allow an unambiguous
determination of the two phases, as for both the Sb:Zn ratio
was close to 1:1.
Figure 5 shows projections of the reconstructed reciprocal
volume from ADT data for orthorhombic phase I. The unit cell
parameters agree well with the known structure of ZnSb (Table
1).11 The systematic extinctions are compatible with the space
group Pbca. Lattice parameters determined directly from ADT
data, as well as subsequently refined against powder X-ray
diffraction data, are approximately 4% larger than those reported
for the ZnSb bulk phase.
The most important information about the ADT data set used
for structure analysis of phase I is compiled in Table 2. The
structure was solved ab initio in one step using direct methods
implemented in SIR200841 from ADT data without electron
beam precession (Figure SI-8, Supporting Information). A fully
High-resolution TEM measurements of the final product as
well as electron diffraction patterns clearly revealed the presence
of a crystallographic phase different from any known Zn4Sb3
phase. Although powder X-ray profiles could be fitted using
the crystallographic data of Zn6Sb5,11 TEM delivered a pseudohex-
agonal lattice and interplanar distances which could not be
(41) Sheldrick, G. M. Acta Crystallogr., Sect. A 2008, 64, 112–122.
9
9884 J. AM. CHEM. SOC. VOL. 132, NO. 28, 2010

Synthesis of a Thermoelectric Zn1+xSb Nanophase
A R T I C L E S
Figure 5. Projections of the full 3D reciprocal space for the orthorhombic phase I of the data set I.1 exhibiting the systematic extinctions of space group
Pbca (see Figure SI-7, Supporting Information, and comment). The volume was visualized using the ADT-3D package applying a volume-rendering technique
with manually determined opacity transfer functions.
Table 1. Crystallographic Data of the Two Different Crystalline
Phases
ZnSb
Zn1+δSb
cell parameters
ADT
refined XRPD
11
ADT
XRPD
space group
a (Å)
b (Å)
Pbca
6.54
8.06
8.31
90
P-1
6.46
8.11
8.43
89.31
89.86
89.38
6.20
7.74
8.10
90
90
90
15.31
15.51
7.81
88.87
89.42
119.4
15.25
15.72
7.81
90
90
120
c (Å)
R (°)
ꢀ (°)
90
90
γ (°)
Table 2. Summary of the ADT Data Sets Employed for Structure
Determination
ZnSb
Zn1+δSb
data set
I.1
II.1
II.2
data type
tilt range
ADT
60°
ADT + PED
76° + 76°
6271
3651
0.8
57
16.86
0.030
36.19
ADT + PED
90°
total reflections
independent reflections
resolution (Å)
reflection coverage, %
Rsym, % of the data set
U isotropic (Å2)
final R, %
388
4486
2745
0.8
106
1.1
70
30.60
0.037
27.10
44
17.45
0.031
42.52
kinematical approach was used (intensities proportional to Fhkl2),
and no correction was applied to the data. Four potential peaks
were identified by the program. The most intense peak (the
height of 1596) was assigned to an Sb atom; the second peak
(the height of 910) was assigned to Zn. The ratio between the
heights of these two peaks corresponds nicely to the ratio
between the atomic numbers of the two atoms 51:30, which in
turn reflects the maximal amplitudes of the scattering factors
for electrons. The other two peaks found by SIR2008 were much
lower (305 and 300) and therefore were treated as ghosts. The
structure model exhibits a close match to the known crystal
structure of ZnSb11 with the errors in atomic positions of 0.02
Å for Sb and 0.05 Å for Zn.
Figure 6. Electron diffraction data for phase II (Zn1+δSb): (a) projection
of the reconstructed reciprocal space along the a* direction; (b) projection
of the reconstructed reciprocal space along the c* direction; (c) single PED
pattern zone [100]; (d) single PED pattern zone [001]; (e, f) intensity scans
of patterns shown in (c) and (d) (line marked by arrows).
To determine the correct crystal system for phase II, single
zone diffraction patterns taken at optimized electron beam
precession conditions were recorded (Figure 6c,d). Both, zone
[100] and zone [001] exhibit strong pseudo-symmetry. However,
a careful inspection of the [100] zone shows that the only true
symmetry element present is the inversion center (due to
Friedel’s law). Figure 6e shows an intensity scan along the row
marked in Figure 6c by dotted arrows: these reflections do not
have mirror symmetry along the [010] direction. The [001] zone
shows no mirror symmetry (Figure 6f), and a careful analysis
of prominent spots, such as, for example, (400), (040), and
(-440), leads to an exclusion of a 3-fold axis. Therefore, the
true zone symmetry is inversion only, and the structure is
triclinic.
Main projections of a 3D reconstructed data set recorded for
phase II (Zn1+δSb) are shown in Figure 6a,b. The corresponding
lattice parameters are listed in Table 1. ADT data confirm that
the cell parameters did not match those of any known Zn-Sb
phase, even if the new phase has to be strictly geometrically
related with rombohedral Zn6Sb5.11 No systematic extinction
was detected. The lattice parameters obtained from electron
diffraction allowed reasonable le Bail fitting of the powder X-ray
diffraction profile, taking into account impurities of Sb and ZnSb
(Table 1).
For structure analysis of phase II, two independent tilt series
were collected. The first crystal was tilted in a range of 76°,
rotated in-plane by approximately 90° and tilted again by 76°
collecting an additional tilt series in the same tilt range. These
9
J. AM. CHEM. SOC. VOL. 132, NO. 28, 2010 9885

A R T I C L E S
Birkel et al.
two tilt series acquired from the same crystal were subsequently
merged (merging factors 1 to 1) and gave data sets II.1 (Table
2). The second data set II.2 was collected from another crystal
within a tilt range of 90° (Table 2). The diffraction camera length
was reduced to achieve a resolution of 0.8 Å. Both data sets
include around 3000 independent reflections.
Crystal Structure of Zn1+δSb. Projections of the crystal
structure are presented and compared with Zn13Sb1013 in Figure
7. Single zone diffraction patterns of phase II (Figure 5c and d)
show a strong superstructure imprint with hexagonal symmetry.
In fact, the Sb substructure displays 6/m symmetry with all Sb
placed on special positions according to the 2e, 2d, 6i, 6j, 6j,
6k Wyckoff sites of space group P6/m. The Zn substructure
reduces the symmetry of the complete structure to -1. Yet, parts
of the Zn substructure exhibit some higher symmetry: i.e., some
of the nonsymmetry equivalent coordination polyhedra are very
similar. For example, the polyhedrons in the middle row
arranged along [010] show very similar organization (cf. Figure
7a). This result is quite impressive as no specific symmetry
constraints were imposed during the structure model refinement.
The structure can be classified within the Zintl-Klemm
concept.44-46 Antimony being more electronegative than Zn
forms two types of anions: isolated Sb3- anions and (Sb2)4-
dumbbells (molecule-anions). The Sb3- anions (Sb1, Sb2, Sb3,
and Sb7) in Zn1+δSb form a flat hexagonal net at z ) 1/2 with
the distances of 4.4-4.6 Å between the nodes (Figure 8a). This
net is placed at 1/2 height of the unit cell (in [001] direction).
The remaining Sb species belong to two different kinds of
(Sb2)4- dumbbells. The first ones are those aligned in [001]
direction and placed with the centers of gravity at virtually the
same height as the hexagonal net (Figure 8b). These antimony
species are located at four corners of the unit cell (Sb9), at four
faces (Sb6, Sb8), and in the center (Sb10). In total, there are
four (Sb2)4- dumbbells of this kind (Figure b). The second type
of dumbbells (Sb4, Sb5, Sb11, Sb12, Sb13, Sb14) is placed at
the z ) 0 and is aligned within the (001) plane in a windmill
pattern (Figure 8c).
For phase II, structure solution by direct methods based on
standard ADT data did not result in a chemically meaningful
structure model. The ADT-PED coupled data set should be
superior for structure solution;35 therefore, as a next step these
data were tried. Indeed, a first solution in space group P1
converged to a reasonable structural model in one step. A check
of the atomic positions displayed immediately the presence of
an inversion center in the structure. This additional symmetry
element was also evident from the correlation coefficient matrix
in the initial structure model refinement. Therefore the structure
was finally solved in space group P-1. A fully kinematical
approach was used, and no correction was applied to the data
set. The correct solution was well recognizable as it exhibited
by far the lowest final residual factor and the highest figure of
merit at the same time. The first 30 maxima in the electron
density map determined by SIR200841 showed reasonable
interatomic distances. The 14 strongest maxima were attributed
to Sb and the remaining 16 to Zn. The structure was subse-
quently refined using SHELXL97,42 showing an appreciable
stability without any constraints. The final refinement converges
at R ) 0.362 and 0.425 for data I and II, respectively, which is
in the typical level of quality for crystal structure analysis by
ADT and electron diffraction in general.35 As the Zn4Sb3
compounds are known to exhibit occupational disorder of the
Zn partial structure, the atomic occupancies for all atoms were
tentatively refined based on the available electron diffraction
data. Even if the reliability of site occupation factors derived
from electron diffraction data is a subject of discussion, values
significantly lower than 1 are usually taken as a hint toward
occupational disorder.43 Taking into account the partial oc-
cupancies, the final stoichiometry of the Zn1+δSb phase is
Several crystalline phases are known within the Zn-Sb phase
diagram close to the 1:1 composition: orthorhombic ZnSb
(Pbca), Zn13Sb10 (three polymorphs crystallizing in space groups
R-3c, C2/c, and C-1 (11)), and R- and ꢀ-Zn4Sb3.14,15 The
structure of Zn1+δSb is very closely related to the rhombohedral
polymorph of Zn13Sb10: both display a hexagonally packed
channel structure. This similarity is seen the best in [001]
projections (Figure 7a and b). The lattice parameters a and b
are connected by the transformation sketched in Figure 7:
aP-1 ) 2/3 × (aR3c - bR3c), bP-1 ) 2/3 × (aR3c + 2bR3c). Yet, the
two structures look very different when viewed from the side
(Figure 7c and d). In the rhombohedral polymorph of Zn13Sb10,
the Sb24- dumbbells are shifted up and down along [001], while
the Sb24- dumbbells in Zn1+δSb are located on the same height.
The orthorhombic ZnSb, in turn, has a more complicated
arrangement consisting of three independent intercalating sets
Zn28.1Sb26.3
.
A full occupancy of the positions results in a density of 5.63
g/cm3, which is in good agreement with the density of 5.68
g/cm3 calculated for the orthorhombic ZnSb phase. If the refined
occupancies are taken into account, the density drops to 5.10
g/cm3. A possible explanation is that the vacancies are partially
compensated by interstitials as often found for such com-
pounds.11 Final residual electron density maps still displayed
intermediate to weak maxima which point to further Zn
positions. Yet, within the given level of detail of electron
diffraction an expansion of the structure model, taking these
maxima into account, would not be appropriate. Albeit the
occupational disorder of the Zn partial substructure, the isotropic
displacement factors were refined stable and exhibited reason-
able values (Table SI-2, Supporting Information).
4-
of Sb2 dumbbells.
In contrast to the rhombohedral polymorph of Zn13Sb10, in
which each Sb atom of a dumbbell is coordinated by three Zn
atoms, the Sb atoms in Zn8Sb7 are coordinated by four Zn atoms
(compare Figure 7c and d). This coordination of Sb is, in turn,
consistent with the coordination observed in ZnSb.
To confirm the above model, the second ADT+PED data
set II.2 was analyzed using the same routine. The result confirms
the structure model described above. The maximal difference
of the atomic positions was 0.41 Å, and the average difference
was 0.18 Å.
The distances between Sb atoms within the dumbbells range
from 2.72 to 2.91 Å and are in a good agreement with the
corresponding distances for the R-3c Zn13Sb10 structure and for
orthorhombic ZnSb. The distances between the Sb in the
dumbbells and the coordinating Zn also agree well with the
(42) Burla, M. C.; Caliandro, R.; Camalli, M.; Carozzini, B.; Cascarano,
G. L.; De Caro, L.; Giacovazzo, C.; Polidori, G.; Siliqi, D.; Spagna,
R. J. Appl. Cryst. 2007, 40, 609-613.
(43) Jansen, J. Structure Refinement by Taking Dynamical Diffraction into
Account; Springer: Netherlands, 2006.
(44) Zintl, E. Angew. Chem. 1939, 52, 1–6.
(45) Nesper, R. Angew. Chem. 1991, 103, 805-834; Angew. Chem., Int.
Ed. 1991, 30, 789-817.
(46) Papoian, G. A.; Hoffmann, R. Angew. Chem. 2000, 112, 2500-2544;
Angew. Chem., Int. Ed. 2000, 39, 2409-2448.
9
9886 J. AM. CHEM. SOC. VOL. 132, NO. 28, 2010

Synthesis of a Thermoelectric Zn1+xSb Nanophase
A R T I C L E S
Figure 7. Relation between the crystal structure of Zn1+δSb and Zn13Sb10.13 (a) [001] view of the structure solved in the previous work; (b) [001] view of
Zn13Sb10; (c) coordination motives of the found structure; (d) coordination motives of Zn13Sb10.
range from 2.64 to 3.0 Å, and are 2.68 Å for the rhombohedral
polymorph of Zn13Sb10.
All structures are related and include common motifs. Under
nonequilibrium conditions, it is difficult to predict which
crystalline phase forms; the output of the reaction is ruled by
kinetic effects rather than the Gibbs free energy. In contrast to
the solid state synthesis of nano-Zn4Sb3 reported by Schlecht,29
this wet chemistry approach produces a Zn4Sb3 related nanophase
in very short reaction times. This is due to the high collision
rate of individual particles in liquid dispersion. On the other
hand, this reaction triggered by nanoparticle collisions leads to
a comparably broad particle size distribution and low crystal-
Figure 8. Arrangement of different Sb species in Zn1+δSb. (a) Hexagonal
arrangement of Sb3-; (b) (Sb2)4- dumbbells orthogonal to the (001) plane;
(c) windmill arranged (Sb2)4- dumbbells in the (001) plane.
linity of the particle surfaces.
Conclusion
known structures: for the newly found structure these distances
range from 2.66 to 3.07 Å, for the orthorhombic ZnSb phase
In conclusion, we have demonstrated a new low-temperature
approach to the synthesis of atomically ordered nanocrystals
9
J. AM. CHEM. SOC. VOL. 132, NO. 28, 2010 9887

A R T I C L E S
Birkel et al.
In a typical synthesis, nanoparticles of the nominal composition
Zn4Sb3 were prepared by heating of Sb nanoparticles (0.275 mmol)
with a 9-fold excess of Zn particles (2.475 mmol) in trioctylamine
at ∼300 °C for 225 min. The resulting black product was collected
by centrifugation (9000 rpm), washed with ethanol, and dried under
a steady Ar flow. Excess of zinc was removed by repeated treatment
of the particles with 2% acetic acid for about 10 min. The product
was then washed with ethanol and dried with streaming Ar to a
dry powder.
X-ray Powder Diffraction. X-ray powder diffraction data were
collected with a Bruker-AXS D8-Discover diffractometer in reflec-
tion geometry equipped with a HiStar detector using graphite
monochromatized Cu KR radiation. Samples were glued on top of
glass and (111) silicon substrates, respectively, using VP/VA
copolymer (vinylpyrrolidone/vinylacetate). le Bail fits and Rietveld
refinements were performed with TOPAS Academic V1.047 ap-
plying the fundamental parameter approach.
TEM and ADT. For TEM investigations, the sample was
suspended in ethanol and sprayed onto a carbon-coated copper grid
using the sonifier described in ref 35. The TEM work was carried
out with a Tecnai F30 S-TWIN transmission electron microscope
equipped with a field emission gun working at 300 kV. High-
resolution (HR-) TEM and electron diffraction patterns were
acquired with a CCD camera (14-bit GATAN 794MSC). EDX
spectra were recorded in STEM mode and quantified in Emispec
ESVision software.
using nanoparticle precursors. By utilizing reduced nanosize
metal precursors, solid-solid diffusion is not rate limiting;
therefore, low reaction temperatures can be used that allow the
formation of phases in the subsolidus regime. Avoiding high-
temperature annealing minimizes sintering, and the crystalline
nanoparticles can be redispersed in solution for high-volume
applications. Zn1+δSb could be obtained only in the presence
of a large excess of Zn that was removed subsequently by mild
acid treatment. Since diffusion clearly precedes nucleation in
the bimetallic precursors, it is possible to synthesize ordered
intermetallic nanocrystals near 200 °C without a high-temper-
ature annealing step.
The TEM data show the presence of at least two crystalline
phases. This raises the fundamental question how powerful and
reliable structure determinations based on X-ray powder dif-
fraction for nanomaterials with complex structures and mixtures
really are. Further insight was obtained using high-resolution
electron diffraction analysis of single nanoparticles. In all
samples, we observe two crystalline phases, ZnSb and Zn1+δSb,
phase segregated from the nano-Zn4Sb3 precursor that was
obtained in the primary reaction from activated Zn and Sb
nanoparticles.
X-ray powder diffraction data could be fitted in a first step
to the crystallographic phases Zn4Sb3 and ZnSb, while they do
not allow the distinction between the Zn4Sb3 and Zn1+δSb phase
due to the qualitity of the XRD data, the profile broadening,
and the complexity of the corresponding crystal structures.
Anyway, HRTEM showed immediately that most of the sample
consisted of a new phase that cannot be fit inside the Zn-Sb
system. A deeper insight was obtained using electron diffraction
tomography of several single crystalline nanoparticles, obtaining
cell parameters and space group of known ZnSb and a new
Zn1+δSb phase. Ab initio structure solutions were performed
for both the phases on the basis of electron diffraction data
obtained by automated diffraction tomography (ADT). Whereas
for the orthorhombic ZnSb structure a rather limited nonprec-
essed ADT data set was sufficient to obtain a structure solution,
precession electron diffraction data were necessary for the more
complex phase Zn1+δSb. This new Zn1+δSb phase crystallizes
with a hexagonal pseudosymmetry in the triclinic space group
P-1 and shows strict relations with both Zn13Sb10 and ZnSb.
Remarkably, the structure was obtained by single-crystal analysis
of a 50 nm particle, and the solution was achieved ab initio in
one step with a fully kinematical approach. The correctness of
the structure is supported by the reproducibility of the solution
and the stability of Fourier and least-squares refinements.
ADT data acquisition was performed with a FISCHIONE
tomography holder, using the ADT acquisition module described
in ref 30. STEM images and diffraction patterns were collected
using a mild illumination setting resulting in an electron dose rate
of 10-15 e/Å2s. STEM images were collected by a FISCHIONE
high angular annular dark field detector (HAADF). Nano electron
diffraction was performed employing a 10 µm C2 condenser
aperture with a 50 nm beam on the sample. ADT data were collected
in standard ED as well as PED mode in steps of 1°.
PED was performed using the SpinningStar unit NanoMEGAS.48
The precession angle was kept at 0.6°, and the precession frequency
was 100 Hz.
The self-developed ADAP package31 written in Matlab was used
for the data processing, including 3D diffraction volume reconstruc-
tion, automated cell parameter determination, and intensity integra-
tion procedures. For the visualization and validation of the 3D data
sets, we used the newly developed ADT-3D software package
programmed in C++.49 The ab initio structure solution was
performed by direct methods implemented in SIR2008, included
in the package Il Milione.41 The structure refinement was performed
using SHELXL97.42
Acknowledgment. This work was supported by the Deutsche
Forschungsgemeinschaft (DFG) within the priority program 1386
Nanostrukturierte Thermoelektrika: Theorie, Modellsysteme und
kontrollierte Synthese and the Deutsche Forschungsgemeinschaft
(DE 412/37-1). C.S.B. is a recipient of a fellowship from MATCOR,
the Graduate School of Excellence of the State of Rhineland-
Palatinate. We acknowledge the help of Mr. G. Glasser with the
SEM and support from the Materials Science Center (MWFZ) in
Mainz. We thank Ulrich Heil, Sebastian Schlitt, and Prof. Elmar
Scho¨mer from the Institute of Computer Science, Johannes Guten-
berg-University of Mainz, for cooperation evolving the ADT-3D
Experimental Section
Synthesis. ZnCl2 (ABCR) was first dried with thionylchloride,
washed with toluene, dried in vacuum, and stored in a glovebox
under N2 atmosphere. Tetrahydrofuran was dried with CaCl2 and
Na/K and freshly distilled before use. Lithium triethylboronhydride
Li[Et3BH] (Aldrich, 1 M in THF, referred to as “superhydride”)
and SbCl3 (ABCR) were used as obtained, and SbCl3 was stored
in a glovebox. Trioctylamine (Aldrich) was degassed and stored
under Ar before use.
Zn particles were synthesized by reaction of ZnCl2 with 2 equiv
of the 1 M solution of “superhydride” in THF at ∼65 °C. The
particles were washed several times with THF, dried in a vacuum,
and stored in a glovebox. Sb particles were produced by reducing
SbCl3 with 3 equiv of the 1 M “superhydride” solution at room
temperature. The black particles were repeatedly redispersed in THF
and decanted from the solution, dried in a vacuum, and kept in a
glovebox.
(47) Coelho, A. TOPAS Academic V1.0; Coelho Software: Brisbane,
Australia, 2004.
(48) NanoMEGAS, AdVanced Tools for Electron Diffraction; NanoMEGAS:
Brussels, Belgium, 2004.
(49) Heil, U.; Schlitt, S.; Kolb, U.; Gorelik, T.; Mugnaioli, E.; Scho¨mer,
E. ADT-3D. A software package for ADT data Visualizing and
processing, Cooperation with the Institute of Computer Science,
Johannes Gutenberg University of Mainz: Mainz, Germany, 2009.
9
9888 J. AM. CHEM. SOC. VOL. 132, NO. 28, 2010

Synthesis of a Thermoelectric Zn1+xSb Nanophase
A R T I C L E S
software package used for 3d data processing and visualization.
Dedicated to Prof. Robert Schöllhorn on the occasion of his 75th
birthday.
EDX data for Zn1+δSb, (7) projection of the full 3D reciprocal
space for the orthorhombic phase I, (8) plot of the ZnSb
structure, (9) atomic coordinates of Zn and Sb for ZnSb
obtained using nonprecessed electron diffraction data, and
(10) atomic coordinates of Zn and Sb for Zn1+δSb. This
material is available free of charge via the Internet at http://
pubs.acs.org.
Supporting Information Available: (1) 1H NMR data of Sb,
(2) DTA data of the Zn1+xSb sample showing the exothermic
decomposition of the Zn1+δ phase (∼200 °C) and the endo-
thermic melting of Zn (∼410 °C), (3) X-ray diffraction and
le Bail fits of the raw product, (4) SEM overview image of
Zn1+xSb, (5) HRTEM image of a Zn1+δSb nanoparticle, (6)
JA1035122
9
J. AM. CHEM. SOC. VOL. 132, NO. 28, 2010 9889